CrUX History API

The CrUX History API gives low-latency access to six months of historical real-user experience data at page and origin granularity.

Try it!

Common use case

The CrUX History API allows for the querying of historical user experience metrics for a specific URI like "Get the historical UX trends for the https://example.com origin."

The History API follows the same structure as the daily CrUX API except values are given in an array, and keys are labelled with plural names (for example, histogramTimeseries instead of histogram, or p75s instead of p75).

CrUX API Key

Like the daily API, using the CrUX History API requires a Google Cloud API key provisioned for Chrome UX Report API usage.. The same key can be used for the daily and history API.

Acquiring and using an API key

Or create one in the Credentials page.

After you have an API key, your application can append the query parameter key=yourAPIKey to all request URLs.

The API key is safe for embedding in URLs; it doesn't need any encoding.

See Example queries.

Data model

This section details the structure of data in requests and responses.

Record

A discrete piece of information about a page, or site. A record can have data that is specific for an identifier and for a specific combination of dimensions. A record can contain data for one or more metrics.

Identifiers

Identifiers specify what records should be looked up. In CrUX these identifiers are webpages and websites.

Origin

When the identifier is an origin all data present for all pages in that origin are aggregated together. For example, say the http://www.example.com origin had pages as laid out by this sitemap:

http://www.example.com/
http://www.example.com/foo.html
http://www.example.com/bar.html

This would mean that when querying the Chrome UX Report with the origin set to http://www.example.com, data for http://www.example.com/, http://www.example.com/foo.html, and http://www.example.com/bar.html would be returned, aggregated together, because those are all pages under that origin.

URLs

When the identifier is a URL, only data for that specific URL will be returned. Looking again to the http://www.example.com origin sitemap:

http://www.example.com/
http://www.example.com/foo.html
http://www.example.com/bar.html

If the identifier is set to URL with the value of http://www.example.com/foo.html, only data for that page will be returned.

Dimensions

Dimensions identify a specific group of data that a record is being aggregated against. For example, a form factor of PHONE indicates that the record contains information about loads that took place on a mobile device.

The CrUX History API is only available aggregated by form factor dimension. This is a general class of device split into PHONE, TABLET, and DESKTOP.

Metric

We report metrics in timeseries of statistical aggregations, which are histograms, percentiles, and fractions.

Histograms

When metrics are expressed in a histogram array, then each timeseries entry represents the percentage of page loads for which the metric fell into an interval, proportionally to all. The data points are presented in the order of the collection period dates also returned by the API, with the first point being the earliest period, and the final point being the most recent collection period.

A three bin histogram for an example metric looks like this:

{
  "histogramTimeseries": [
    {
      "start": 0,
      "end": 2500,
      "densities": [0.9190, 0.9203, 0.9194, 0.9195, 0.9183, 0.9187]
    },
    {
      "start": 2500,
      "end": 4000,
      "densities": [0.0521, 0.0513, 0.0518, 0.0518, 0.0526, 0.0527]
    },
    {
      "start": 4000,
      "densities": [0.0288, 0.0282, 0.0286, 0.0285, 0.0290, 0.0285]
    }
  ],
}

This data indicates that 91.90% of page loads experienced the example metric value between 0ms and 2,500ms for the first collection period in the history, followed by 92.03%, 91.94%... The units of the metric are not contained in this histogram, in this case we will assume milliseconds.

Additionally, 5.21% of page loads experienced the example metric value between 2,500ms and 4,000ms in the first collection period in the history, and 2.88% of page loads experienced a value greater than 4,000ms in the first collection period in the history.

Percentiles

Metrics may also contain timeseries of percentiles that can be useful for additional analysis.

The data points are presented in the order of the collection period dates also returned by the API, with the first point being the earliest period, and the final point being the most recent collection period.

{
  "percentilesTimeseries": {
    "p75s": [1362, 1352, 1344, 1356, 1366, 1377]
  },
}

These percentiles can show specific metric values at the given percentile for that metric. They are based on the full set of available data and not the final binned data, so they don't necessarily match an interpolated percentile that is based on the final binned histogram.

Fractions

Metrics may be expressed as timeseries of labeled fractions; each label describes a page load in a particular way. The data points are presented in the order of the collection period dates also returned by the API, with the first point being the earliest period, and the final point being the most recent collection period.

Example:

{    
  "fractionTimeseries": {
    "desktop": {"fractions": [0.3195, 0.2115, 0.1421]},
    "phone": {"fractions": [0.6295, 0.7544, 0.8288]},
    "tablet": {"fractions": [0.051, 0.0341, 0.029]}
  }
}

In this example, the most recent data point indicates 14.21% of page loads came from desktop, and 82.88% came from phones.

Metric value types

As the CrUX History API uses the same metric value types, you can reference the daily CrUX API metric value types documentation for more details.

Metric eligibility

Based on the eligibility criteria an origin or URL may only be eligible for some of the collection periods covered by the CrUX History API. In these cases the CrUX History API will return "NaN" for the histogramTimeseries densities and null for the percentilesTimeseries for the collection periods which have no eligible data. The reason for the difference is the histogram densities are always numbers, while the percentiles can be numbers or strings (CLS uses strings, even if they look like numbers).

For example, if the second period did not have any eligible data, this would show as:

{
  "histogramTimeseries": [
    {
      "start": 0,
      "end": 2500,
      "densities": [0.9190, "NaN", 0.9194, 0.9195, 0.9183, 0.9187]
    },
    {
      "start": 2500,
      "end": 4000,
      "densities": [0.0521, "NaN", 0.0518, 0.0518, 0.0526, 0.0527]
    },
    {
      "start": 4000,
      "densities": [0.0288, "NaN", 0.0286, 0.0285, 0.0290, 0.0285]
    }
  ],
  "percentilesTimeseries": {
    "p75s": [1362, null, 1344, 1356, 1366, 1377]
  },
}

For URLs or origins that fall in and out of eligibility over time, you may notice many missing entries.

Collection periods

The CrUX History API contains a collectionPeriods object with an array of firstDate and endDate fields representing the beginning and end dates of each aggregation window. For example:

    "collectionPeriods": [{
        "firstDate": { "year": 2022, "month": 7, "day": 10 },
        "lastDate": { "year": 2022, "month": 8, "day": 6 }
      }, {
        "firstDate": { "year": 2022, "month": 7, "day": 17 },
        "lastDate": { "year": 2022, "month": 8, "day": 13 }
      }, {
        "firstDate": { "year": 2022, "month": 7, "day": 24 },
        "lastDate": { "year": 2022, "month": 8, "day": 20 }
      }, {
        "firstDate": { "year": 2022, "month": 7, "day": 31 },
        "lastDate": { "year": 2022, "month": 8, "day": 27 }
      }, {
        "firstDate": { "year": 2022, "month": 8, "day": 7 },
        "lastDate": { "year": 2022, "month": 9, "day": 3 }
      }, {
        "firstDate": { "year": 2022, "month": 8, "day": 14 },
        "lastDate": { "year": 2022, "month": 9, "day": 10 }
      }
    ]

These collection periods are in ascending order and represent the date span of each of the data points in the other sections of the response.

The History API is updated each Monday and contains data up until the previous Saturday (as per the standard 2-day lag). It contains the previous 25-weeks worth of data—one collection period per week.

As each collection period contains the previous 28-days aggregated data, and the collection periods are per week, this means the collection periods will overlap. They are similar to a moving average of data, with three weeks worth of data being included in each subsequent period, and one week being different.

Example queries

Queries are submitted as JSON objects using a POST request to https://chromeuxreport.googleapis.com/v1/records:queryHistoryRecord?key=[YOUR_API_KEY]" with query data as a JSON object in the POST body.

Note the use of queryHistoryRecord replacing the queryRecord of the daily CrUX API.

An example body is:

{
  "origin": "https://example.com",
  "formFactor": "PHONE",
  "metrics": [
    "largest_contentful_paint",
    "experimental_time_to_first_byte"
  ]
}

For example, this can be called from curl with the following command line (replacing API_KEY with your key):

curl -s --request POST 'https://chromeuxreport.googleapis.com/v1/records:queryHistoryRecord?key=API_KEY' \
    --header 'Accept: application/json' \
    --header 'Content-Type: application/json' \
    --data '{"formFactor":"PHONE","origin":"https://www.example.com","metrics":["largest_contentful_paint", "experimental_time_to_first_byte"]}'

Page-level data is available through the API by passing a url property in the query, instead of origin:

{
  "url": "https://example.com/page",
  "formFactor": "PHONE",
  "metrics": [
    "largest_contentful_paint",
    "experimental_time_to_first_byte"
  ]
}

If the metrics property is not set then all available metrics will be returned:

• cumulative_layout_shift
• first_contentful_paint
• interaction_to_next_paint
• largest_contentful_paint
• experimental_time_to_first_byte
• navigation_types
• round_trip_time
• form_factors (only reported if no formFactor is specified in the request)

If no formFactor value is provided then the values will be aggregated across all form factors.

See Using the CrUX History API guide for more example queries.

Data pipeline

The CrUX dataset is processed through a pipeline to consolidate, aggregate and filter the data before becoming available through the API.

The rolling average

The data in the Chrome UX Report is a 28-day rolling average of aggregated metrics. This means that the data presented in the Chrome UX Report at any given time is actually data for the past 28 days aggregated together.

The History API contains a number of collection periods, each spanning these 28 days. As each collection period contains the previous 28-days aggregated data, and the collection periods are per week, this means the collection periods will overlap. They are similar to a moving average of data, with three weeks worth of data being included in each subsequent period, and one week being different.

Weekly updates

The History API is updated each Monday around 04:00 UTC and contains data up until the previous Saturday (as per the standard 2-day lag). It contains the previous 25 weeks (approximately 6 months) worth of data, one collection period per week.

There is no service level agreement for update times; it is run on a best-effort basis every day.

Schema

There is a single endpoint for the CrUX History API which accepts POST HTTP requests. The API returns a record which contains one or more metrics corresponding to performance data about the requested origin or page.

HTTP request

POST https://chromeuxreport.googleapis.com/v1/records:queryHistoryRecord

The URL uses gRPC Transcoding syntax.

Request body

The CrUX History API uses the same request bodies as the daily CrUX API.

For example, to request the desktop Largest Contentful Paint values for the web.dev homepage:

{
  "origin": "https://web.dev/",
  "formFactor": "DESKTOP",
  "metrics": [
    "largest_contentful_paint"
  ]
}

Response body

Successful requests return responses with a record object and urlNormalizationDetails in the following structure:

{
  "record": {
    "key": {
      object (Key)
    },
    "metrics": [
      string: {
        object (Metric)
      }
    ]
  },
  "urlNormalizationDetails": {
    object (UrlNormalization)
  }
}

For example, the response to the request body in the previous request could be:

{
  "record": {
    "key": {
      "origin": "https://web.dev"
    },
    "metrics": {
      "largest_contentful_paint": {
        "histogramTimeseries": [{
            "start": 0, "end": 2500, "densities": [
              0.9190, 0.9203, 0.9194, 0.9195, 0.9183, 0.9187, ...
            ]
          }, {
            "start": 2500, "end": 4000, "densities": [
              0.0521, 0.0513, 0.0518, 0.0518, 0.0526, 0.0527, ...
            ]
          },  {
            "start": 4000, "densities": [
              0.0288, 0.0282, 0.0286, 0.0285, 0.0290, 0.0285, ...
            ]
          }
        ],
        "percentilesTimeseries": {
          "p75s": [
            1362, 1352, 1344, 1356, 1366, 1377, ...
          ]
        }
      }
    },
    "collectionPeriods": [{
        "firstDate": { "year": 2022, "month": 7, "day": 10 },
        "lastDate": { "year": 2022, "month": 8, "day": 6 }
      }, {
        "firstDate": { "year": 2022, "month": 7, "day": 17 },
        "lastDate": { "year": 2022, "month": 8, "day": 13 }
      }, {
        "firstDate": { "year": 2022, "month": 7, "day": 24 },
        "lastDate": { "year": 2022, "month": 8, "day": 20 }
      }, {
        "firstDate": { "year": 2022, "month": 7, "day": 31 },
        "lastDate": { "year": 2022, "month": 8, "day": 27 }
      }, {
        "firstDate": { "year": 2022, "month": 8, "day": 7 },
        "lastDate": { "year": 2022, "month": 9, "day": 3 }
      }, {
        "firstDate": { "year": 2022, "month": 8, "day": 14 },
        "lastDate": { "year": 2022, "month": 9, "day": 10 }
      }, {
        ...
      }
    ]
  }
}

Key

Key defines all the dimensions that identify this record as unique.

{
  "formFactor": enum (FormFactor),

  // Union field url_pattern can be only one of the following:
  "origin": string,
  "url": string
  // End of list of possible types for union field url_pattern.
}

Metrics

A metric is a set of user experience data for a single web performance metric, such as first contentful paint. It contains a summary histogram of real world Chrome usage as a series of bins.

{
  "histogramTimeseries": [
    {
      object (Bin)
    }
  ],
  "percentilesTimeseries": {
    object (Percentiles)
  }
}

or

"fractionTimeseries": {
  object (Fractions)
}

Bin

A bin is a discrete portion of data spanning from start to end, or if no end is given from start to positive infinity.

A bin's start and end values are given in the value type of the metric it represents. For example, first contentful paint is measured in milliseconds and exposed as ints, therefore its metric bins will use int32s for its start and end types. However cumulative layout shift is measured in unitless decimals and is exposed as a decimal encoded as a string, therefore its metric bins will use strings for its value type.

{
  "start": value,
  "end": value,
  "densities": [number, number, number...etc.]
}

Percentiles

Percentiles contains synthetic values of a metric at a given statistical percentile. These are used for estimating a metric's value as experienced by a percentage of users out of the total number of users.

{
  "P75": value
}

Fractions

Fractions contains timeseries of labeled fractions that add up to ~1, per entry. Each label describes a page load in some way, so metrics represented in this way can be thought of as producing distinct values instead of numerical values, and the fractions express how frequently a particular distinct value was measured.

{
  "label_1": { "fractions": array[fraction]},
  "label_1": { "fractions": array[fraction]},
  ...
  "label_n": { "fractions": array[fraction]}
}

Much like the density values in histogram bins, each fraction is a number 0.0 <= value <= 1.0, and they add up to ~1.0. When the metric is not available for a particular collection period, then the corresponding entry will be "NaN" in all arrays of fractions.

UrlNormalization

Object representing the normalization actions taken to normalize a URL to achieve a higher chance of successful lookup. These are simple automated changes that are taken when looking up the provided url_pattern would be known to fail. Complex actions like following redirects are not handled.

{
  "originalUrl": string,
  "normalizedUrl": string
}

Rate limits

The CrUX History API shares the same limit with the CrUX API for 150 queries per minute per Google Cloud project for either API, which is offered without charge. This limit, and your current usage, can be seen in the Google Cloud Console. This generous quota should be sufficient for the vast majority of use cases and it is not possible to pay for an increased quota.